JPS62150431A - Rounding designation system - Google Patents

Abstract

PURPOSE:To prevent overhead by selecting any of the 1st designation means designating the kind of rounding and the 2nd designation means designating the kind of rounding through the identification of a prescribed instruction specifically so as to execute the rounding. CONSTITUTION:Each bit of a 4-bit FF 1 set by a write instruction or the like of a CPU corresponds to a round mode, only one bit in the 4-bit is set to have a logical 1 and its value is outputted to a rounding execution circuit 6 according to a priority circuit 5 when the value of a 1-bit FF 4 is 0. Further, when the value of the FF 4 is logical 1, the special instruction is decoded by the instruction decoder 2 and when a line corresponding to signal lines inputted to a PLA 3 is logical 1, the programmed output line is brought into logical 1 and its output is inputted from the PLA 3 to the circuit 6 according to the circuit 5.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to type conversion in a system that performs numerical operations and control of rounding that occurs when operating with different data types.

[Conventional technology]

Numerical performance 9. The types of rounding are 1, for example, IEEE
Now, the rounding mode of the 4 citrons (toward 10 as).

toward -as, toward Q, nea
rest ) is defined. In the conventional numerical performance system, by instructing one of these 4 to 11 rounding types, rounding of one type is executed regardless of the NI class of the number. For example, according to literature (1), the type of rounding is RC (Round -
Control) field allows selection, and when the user specifies one of them.

As long as you do not change the specification, rounding to 1 will be applied to all operations.

Sentence: (1) r IAPX 86 Family User's Manual Chapter 3 Intel Japan (1981) [Issues to be resolved by the invention] The above-mentioned 7 conventional rounding specification methods specify the type of one-inch rounding. Then, only one kind of rounding is not performed. For this reason, it is necessary to execute language input such as C, FORTRAN language, etc. In cases where automatic type conversion and associated rounding are often performed, and where there are several types of rounding methods that correspond to the type of type conversion, different type conversions will occur if only the -Pk class is specified. There is a drawback that the rounding Wtm must be respecified each time, resulting in overhead.

For example, type conversion in C language will be explained using a program example shown in the M3 diagram and calculation results.

1ine 10.17 and 24.31 in the glogram example are type conversions from real numbers to integers, 1ine 11. 18.2
5゜32 is the type conversion of real number from double precision to single precision, and the operation result 1ine2. 5. 8. 11 is the result of a type conversion from a real number to an integer, and 1ine3, 6, 9.12 are the results of a type conversion from double precision to single precision of a real number. As shown in these figures, rounding when converting a real number to an integer is done by rounding, and rounding when converting a real number from double precision to single precision is done by rounding.

Therefore, if you specify another rounding mode during another operation, each time a type conversion occurs, the rounding f11! The class must be respecified.

[Means for solving problems]

The rounding instruction method of the present invention includes a first means for specifying the type of rounding, an identification means for identifying a percentage command, and a second specifying means for specifying the type of rounding according to the instruction of the identification means. , an instruction means for instructing the selection of either the first or the 20th instruction means, and a rounding means for executing rounding according to the instructions of the instruction means, and if there is no instruction by the instruction means, the first one is selected. Rounding is performed according to the specification of the specifying means, and when there is an instruction by the specifying means and a specific instruction is identified by the instruction identifying means, the specification of the 20th specifying means is given priority,
This is a rounding designation method in which rounding is performed, and in other cases, rounding can be performed with priority given to the designation by the first designation means.

〔Example〕

Next, the present invention will be explained with reference to the drawings.

; Figure 1 is a schematic diagram of the configuration of an embodiment of the scope of uninvented poetry, Figure 2 is an explanatory diagram of a priority circuit, and Figure 4 is a correspondence table based on the inclusion of commands and rounding.

In FIG. 1, 1 is a 4-bit flip-flop, 2 is an instruction decoder, 3 is a PLA, 4 is a 1-bit flip-flop, 5 is a priority circuit, and 6 is a rounding execution circuit.

The information of 1 is set by a special insert command of the CPU. Each pit corresponds to a respective rounding mode%
Set the value so that only 1 bit out of 4 bits has a value of 1. If the value of 4 is O, this value is output to 6 according to the priority circuit of 5. Also, if 4's 1st shift is 1, 2
When a specific command is decoded and the corresponding one among the signal lines input to PLA in 3 becomes 1, the output line programmed in PLA becomes 1 according to Table 1, and the priority of 5 becomes 1. It is output from 3 to 6 according to the circuit.

5 is a priority circuit. When the output of 4 is 0, it gives priority to the output from 1 and supplies it to 6. When the output of 4 is 1, if any of the outputs of 3 is 1, output, otherwise give priority to output from 1.

Supply to 6.

Next, a specific example of the sound-light circuit 5 will be explained using the second [Q].

explain. In Fig. 2, 11 to 16 are AND circuits.

17-19 are OR circuits %20 is NAND circuits, 21-24
is the output signal from 1, 25.26 is the output signal from 3,
27 is a 1-bit 7-rip 70g. 28-31 are output signals to 6. For example, assume that the signal 24 from 1 is 1.

If the signal from 27 is 0 at this time, the signal 24 from 1
is output in the evening, and 30 is output to 6.

Next, assume that the signal 24 from 1 is 1 and the signal 27 is 1. At this time, when the signal 25 from 3 is 1, the signal from 3 is taken over and %28 is output to 6. Also, 3
Signal from 25.21? If both are 0, the signal from 1 is skipped and 30 is output to 6.

As explained above, either the signal specified by 1 or the signal specified by 3 is given priority in the output from 4, and rounding is executed.

〔Effect of the invention〕

As explained above, the present invention identifies an instruction in addition to the first designation means that designates the type of rounding.

Providing an instruction means for instructing either the 20th finger/toe means, the first specifying means, or the second specifying means for specifying class f4 rounding by the identification means, and causing rounding to be executed according to the instruction of the instruction means. Therefore, for a specific instruction in which the type of rounding is predetermined, there is no need to respecify rounding every time the inclusion changes, and there is no need for overhead to occur.

[Brief explanation of drawings]

FIG. 1 shows the constituent factors of the implementation f11 of the present invention, FIG. 2 shows an explanatory diagram of the target circuit, and FIG. 3 shows an example of the program. Figure 4 shows the correspondence between commands and rounding. 1...7 rip crops of 4 bits, 2...
...Instruction decoder, 3...PLA, 4...
...7 lip-flops of 1 bit, 5...Priority circuit, 6...Rounding circuit% 11-16...Group/AND circuit, 17-19...Group/OR circuit, 20・・・
Group/Nand circuit, 21-24...output signal from mark 1, 25-26...output signal from 3% 27.
...1 pit 7 lip frog, 28-31.
...Output signal to 6. Figure 7, Figure 3 + 1Hnclude (StdiO,hn3
ma+nH 6float fi 7 ir+t i;9
d mI I 0OOQOOOO
OOOOOOIiIQ f Thoughts
d+16 d=O,'? Q
QQ9q9'? '? Qf9'hZ 23 (1= -l 0000
000000001124
F = d d = -OQQQQ9Q
'? QQ'? QQ9r 3rd ward (anatomy operation cotton inclusion 2 name name 1°5 $I'f!r L ~ n * M
16, Arc rKf1 person hb.

Claims (1)

[Claims] In a system that handles a plurality of data types and performs numerical operations, a first designation means for designating the type of rounding, an identification means for identifying a specific command, and a type of rounding is designated by the instruction from the identification means. a second means, an instruction means for instructing selection of either the first or second specifying means, and a rounding means for executing rounding according to the specification of the instruction means,
If there is no instruction from the instruction means, the rounding means is activated and rounding is performed according to the instruction from the first instruction means, but if there is an instruction from the instruction means and a specific instruction is identified by the instruction identification means. In other cases, the rounding is performed by giving priority to the instruction from the second instruction means, and in other cases, the rounding is performed by giving priority to the instruction from the first instruction means. .